We introduce HypoExplore, an agentic framework that formulates model discovery as a hypothesis-driven scientific inquiry. Given a human-specified high-level research direction, HypoExplore ideates, implements, evaluates, and improves neural architectures through evolutionary branching. Hypotheses are generated using a large language model by selecting a parent hypothesis to build upon, guided by a dual strategy that balances exploiting validated principles with resolving uncertain ones. Our proposed framework maintains a Trajectory Tree that records the lineage of all proposed architectures, and a Hypothesis Memory Bank that actively tracks associated confidence scores acquired through experimental evidence. After each experiment, multiple feedback agents analyze the results from different perspectives and consolidate their findings into hypothesis confidence updates. Under minimal research directions, HypoExplore discovers a model that achieves top results on two medical imaging tasks on MedMNIST, an improved encoder for named entity recognition on FiNER-ORD, and a lightweight 0.7M parameter model on CIFAR-10 that reaches 94.1% accuracy and transfers well to other benchmarks. Our analysis of hypothesis confidence scores across evolutionary lineages suggest that HypoExplore not only discovers stronger architectures, but can help build a genuine understanding of the design space.

Large Reasoning Models (LRMs) are powerful, but they still suffer from inefficient and off-target reasoning. Currently, training-free methods are limited to either rigid heuristics or descriptive, non-actionable analyses. In this paper, we introduce ThinkPilot, a training-free framework that automatically optimizes LRMs reasoning. It uses an evolutionary process to generate think-prefixes, namely instructions that evolve driven by a taxonomy of reasoning behaviors to guide models toward superior performance. Extensive experiments demonstrate ThinkPilot’s broad effectiveness: it significantly improves the accuracy-length trade-off for efficient reasoning, drastically improves safety (e.g., cutting the StrongREJECT score of DeepSeek-R1-Distill-Qwen-32B from 27.0% to 0.7%), and enhances instruction following. It also synergizes with existing training-based methods. Specially, our analysis reveals that think-prefixes can reliably control LRMs’ reasoning behaviors, and that different tasks have strong preferences for specific behavioral distributions. By automatically identifying and eliciting these behaviors, ThinkPilot provides a generalizable framework for aligning LRMs reasoning with task demands.

Streaming speech translation (StreamST) requires determining appropriate timing, known as policy, to generate translations while continuously receiving source speech inputs, balancing low latency with high translation quality. Existing StreamST methods typically operate on sentence-level speech segments, referred to as simultaneous speech translation (SimulST). In practice, they require collaboration with upstream segmentation models to accomplish StreamST, where the truncated speech segments constrain SimulST models to make policy decisions and generate translations based on pre-defined contextual information preset by the upstream models. Moreover, SimulST models struggle to learn effective policies due to the complexity of speech inputs and cross-lingual generation. To address these challenges, we propose StreamUni, which achieves StreamST through a unified Large Speech-Language Model (LSLM). Specifically, StreamUni incorporates speech Chain-of-Thought (CoT) in guiding the LSLM to generate multi-stage outputs. Leveraging these multi-stage outputs, StreamUni simultaneously accomplishes speech segmentation, policy decision, and translation generation. Additionally, we propose a streaming CoT training method that enhances low-latency policy decisions and generation capabilities using limited CoT data. Experiments demonstrate that our approach achieves state-of-the-art performance on both SimulST and StreamST tasks.

Long-context LLMs and Retrieval-Augmented Generation (RAG) systems expand accessible context, but still defer state tracking and evidence consolidation to query time. This passive paradigm is brittle in streaming settings where facts are updated and answers depend on latent states rather than explicit spans. We introduce Unified Memory Agent (UMA), an end-to-end reinforcement-learning framework that trains a single policy to maintain external memory and answer questions. UMA updates a structured memory bank via CRUD operations and combines it with raw-context retrieval for QA. To jointly optimize memory maintenance and question answering, we propose Task-Stratified GRPO, which normalizes heterogeneous trajectories separately for more effective credit assignment. We also introduce Ledger-QA, a controlled benchmark for long-horizon state tracking over accumulated updates. Experiments on Ledger-QA and 13 TTL/AR benchmarks show that UMA improves long-horizon tracking, achieves the best average TTL/AR score among compared methods, and remains competitive on retrieval-centric benchmarks. Our results highlight learned proactive memory management as a complement to longer context windows and query-time retrieval.

Reinforcement Learning with Verifiable Rewards (RLVR) has driven substantial progress in reasoning-intensive domains like mathematics. However, optimizing open-ended generation remains challenging due to the lack of ground truth. While rubric-based evaluation offers a structured proxy for verification, existing methods suffer from scalability bottlenecks and coarse criteria, resulting in a supervision ceiling effect. To address this, we propose an automated Coarse-to-Fine Rubric Generation framework. By synergizing principle-guided synthesis, multi-model aggregation, and difficulty evolution, our approach produces comprehensive and highly discriminative criteria capable of capturing the subtle nuances. Based on this framework, we introduce RubricHub, a large-scale (110k) and multi-domain dataset. We validate its utility through a two-stage post-training pipeline comprising Rubric-based Rejection Sampling Fine-Tuning (RuFT) and Reinforcement Learning (RuRL). Experimental results demonstrate that RubricHub unlocks significant performance gains: our post-trained Qwen3-14B achieves state-of-the-art (SOTA) results on HealthBench (69.3), surpassing proprietary frontier models such as GPT-5.

Despite long-standing efforts in accelerating scientific discovery with AI, building AI co-scientists remains challenging due to limited high-quality data for training and evaluation. To tackle this data scarcity issue, we present AutoSDT, an automatic pipeline that collects high-quality coding tasks in real-world data-driven discovery workflows. AutoSDT leverages the coding capabilities and parametric knowledge of LLMs to search for diverse sources, select ecologically valid tasks, and synthesize accurate task instructions and code solutions. Using our pipeline, we construct AutoSDT-5K, a dataset of 5,404 coding tasks for data-driven discovery that covers four scientific disciplines and 756 unique Python packages. To the best of our knowledge, AutoSDT-5K is the only automatically collected and the largest open dataset for data-driven scientific discovery. Expert feedback on a subset of 256 tasks shows the effectiveness of AutoSDT: 93% of the collected tasks are ecologically valid, and 92.2% of the synthesized programs are functionally correct. Trained on AutoSDT-5K, the Qwen2.5-Coder-Instruct LLM series, dubbed AutoSDT-Coder, show substantial improvement on two challenging data-driven discovery benchmarks, ScienceAgentBench and DiscoveryBench. Most notably, AutoSDT-Coder-32B reaches the same level of performance as GPT-4o on ScienceAgentBench with a success rate of 7.8%, doubling the performance of its base model. On DiscoveryBench, it lifts the hypothesis matching score to 8.1, bringing a 17.4% relative improvement and closing the gap between open-weight models and GPT-4o.

Emotional Intelligence (EI), consisting of emotion perception, emotion cognition and emotion expression, plays the critical roles in improving user interaction experience for the current large language model (LLM) based conversational general AI assistants. Previous works mainly focus on raising the emotion perception ability of them via naive fine-tuning on EI-related classification or regression tasks. However, this leads to the incomplete enhancement of EI and catastrophic forgetting of the general intelligence (GI). To this end, we first introduce EiBench, a large-scale collection of EI-related tasks in the text-to-text format with task instructions that covers all three aspects of EI, which lays a solid foundation for the comprehensive EI enhancement of LLMs. Then a novel Modular Emotional Intelligence enhancement method (MoEI), consisting of Modular Parameter Expansion and intra-inter modulation, is proposed to comprehensively enhance the EI of LLMs without compromise their GI. Extensive experiments on two representative LLM-based assistants, Flan-T5 and LLaMA-2-Chat, demonstrate the effectiveness of MoEI to improving EI while maintain GI.

Although large language models (LLMs) show remarkable capabilities and generalizability across various tasks, they are criticized for lack of expertise. One promising solution is to combine knowledge graphs (KGs) with LLMs, and recent studies focus on integrating KGs into LLMs through prompt-based methods. However, these approaches fail to use the structural information of the KGs, suffer from the problem of knowledge conflict, and over-reliance on super LLMs. To address these challenges, we propose KG-Adapter, a parameter-level KG integration method based on parameter-efficient fine-tuning (PEFT). Specifically, we introduce a novel adapter structure designed for decoder-only LLMs, which can encode KGs from both node-centered and relation-centered perspectives, and then perform joint reasoning with LLMs to generate responses end-to-end. Experiments with diverse models on four datasets for two different tasks all demonstrate significant improvements. With only 28M parameters trained, we make the 7B-parameter LLM outperform the previous full-parameter fine-tuned state-of-the-art method and comparable to the prompt-based ChatGPT methods.

Existing commonsense knowledge bases often organize tuples in an isolated manner, which is deficient for commonsense conversational models to plan the next steps. To fill the gap, we curate a large-scale multi-turn human-written conversation corpus, and create the first Chinese commonsense conversation knowledge graph which incorporates both social commonsense knowledge and dialog flow information. To show the potential of our graph, we develop a graph-conversation matching approach, and benchmark two graph-grounded conversational tasks. All the resources in this work will be released to foster future research.

Existing entity alignment models mainly use the topology structure of the original knowledge graph and have achieved promising performance. However they are still challenged by the heterogeneous topological neighborhood structures which could cause the models to produce different representations of counterpart entities. In the paper we propose a global entity alignment model with gated latent space neighborhood aggregation (LatsEA) to address this challenge. Latent space neighborhood is formed by calculating the similarity between the entity embeddings it can introduce long-range neighbors to expand the topological neighborhood and reconcile the heterogeneous neighborhood structures. Meanwhile it uses vanilla GCN to aggregate the topological neighborhood and latent space neighborhood respectively. Then it uses an average gating mechanism to aggregate topological neighborhood information and latent space neighborhood information of the central entity. In order to further consider the interdependence between entity alignment decisions we propose a global entity alignment strategy i.e. formulate entity alignment as the maximum bipartite matching problem which is effectively solved by Hungarian algorithm. Our experiments with ablation studies on three real-world entity alignment datasets prove the effectiveness of the proposed model. Latent space neighborhood informationand global entity alignment decisions both contributes to the entity alignment performance improvement.

To model diverse responses for a given post, one promising way is to introduce a latent variable into Seq2Seq models. The latent variable is supposed to capture the discourse-level information and encourage the informativeness of target responses. However, such discourse-level information is often too coarse for the decoder to be utilized. To tackle it, our idea is to transform the coarse-grained discourse-level information into fine-grained word-level information. Specifically, we firstly measure the semantic concentration of corresponding target response on the post words by introducing a fine-grained focus signal. Then, we propose a focus-constrained attention mechanism to take full advantage of focus in well aligning the input to the target response. The experimental results demonstrate that by exploiting the fine-grained signal, our model can generate more diverse and informative responses compared with several state-of-the-art models.